Re: [RFA/rs6000-aix] software_single_step gdbarch method no longer set

[Daniel Jacobowitz <drow at false dot org>]

On Sat, Dec 31, 2005 at 09:41:33AM +0400, Joel Brobecker wrote:
> I traced the problem to the following piece of code:
> 
>    if (from_xcoff_exec)
>      {
>        /* NOTE: jimix/2003-06-09: This test should really check for
>          GDB_OSABI_AIX when that is defined and becomes
>          available. (Actually, once things are properly split apart,
>          the test goes away.) */
>         /* RS6000/AIX does not support PT_STEP.  Has to be simulated.  */
>         set_gdbarch_software_single_step (gdbarch, rs6000_software_single_step);
>      }

I have not looked at the patch much (won't this year).  However, if
you're feeling especially motivated, we could sidestep this issue...
there's a related cleanup I've been meaning to do for a long while.

The software single step method in the gdbarch is always used if it is
set.  But, that doesn't make a lot of sense.  The gdbarch method should
mean "software single step is supported"; the predicate "software
single step is required" should live in the target vector instead! 
It's a property of the target.  e.g. MIPS GNU/Linux doesn't implement
PTRACE_SINGLESTEP, but if you're connected to a kernel debugging
session using a hardware ICE, then maybe you've really got hardware
single step.

It's unfortunately a little trickier than that since remote targets
don't currently report whether using software singlestep is OK :-(
So setting the default behavior may not be simple.  I know gdbserver
needs some adjustment here.

Just a thought.

-- 
Daniel Jacobowitz
CodeSourcery, LLC